with considering the possibility of reverse causation among
dependent and independent variables^56.
With these caveats in mind, on the whole, it appears that the
recovery of the business cycle (which is certainly related to
improvements in external conditions), together with the
introduction of pro-poor macroeconomic, labor and social policies
(which is related to the election of LOC regimes) played a major
role, as expected ex ante, in reducing income inequality. Though this
variable moves very slowly, the impact of the distribution of years
of education (which has slowly improved during the last 15 years)
also had an important impact. These results broadly confirm the
theoretical considerations presented in Part 3 regarding the possible
sources of the inequality decline that has taken place in Latin
America in the 2000s. In addition, these results contradict the
conclusions reached by Perez Caldentey and Vernengo (2008)
which state that the recent growth acceleration and fall in
(^56) Reverse causation is tested by means of the Granger test. However, such test is
not suitable for the ADLI dataset in which each variable has, at most, 18 or fewer
observations due to missing data. It is therefore more appropriate to deal with this
problem from a theoretical standpoint. In this regard, it must be noted that reverse
causality makes no sense in the majority of the relations in Table 13. For instance,
it is not plausible that changes in domestic inequality affect the real exchange rate,
or can affect lagged, exogenous or policy variables (such as Gini income 1990,
migrant remittances, terms of trade, ratio of direct/indirect taxes, ratio of pension
coverage Q1/Q5, and minimum wage). Also, a fall/increase in Gini income may
affect the Gini of years of education only after a considerable lag. It is also
implausible that a decline in inequality will affect the expenditure on social
insurance/GDP, which depends on the coverage of formal employment as far as
pensions are concerned, and on tax revenue and public expenditure allocation for
conditional cash transfers. The only area where reverse causation may be plausible
is between the Gini inequality and the growth rate of GDP/c. In this case,
however, this relation would be characterized by time lags, thus excluding the
possibility of reverse causation on synchronous data. Furthermore, the literature
on the impact of higher inequality on GDP/c growth is not unanimous.
Neokeynesian and neoclassical models postulate a positive relation between these
two variables, while ‘political economy’ and ‘incentives’ models assume a negative
one. On the whole, reverse causality does not seem plausible. However, the
parameters in Table 13 may be distorted by the possible endogeneity of some
explanatory variables. Solving formally this problem by means of a simultaneous
equations system is however a difficult task in a panel with 18 countries.